""" Experiment 3: Observer Divergence ----------------------------------- "Does internal coherence correlate with observer-attributed intentionality?" Protocol: 1. Run both agents for N=100 sessions 2. Every 10 sessions: compute (a) agent-internal Δ-Kohärenz and (b) observer intentionality_score 3. Plot divergence between the two over time The scientifically interesting output: Case A: High internal Δ-Kohärenz, Low observer → opaque identity Case B: Low internal, High observer → The Mirror Problem (appears intentional but isn't) Case C: High/High → genuine development visible to observer Case D: Low/Low → baseline mirror behavior """ import os import sys import json import numpy as np import matplotlib.pyplot as plt from datetime import datetime from agents.baseline_mirror_agent import BaselineMirrorAgent from agents.three_layer_agent import ThreeLayerAgent from metrics.delta_coherence import delta_coherence from metrics.observer_attribution import intentionality_score # Reuse topics from exp1 TOPICS = [ "emergence in complex systems", "self-organization and pattern formation", "information theory and entropy", "active inference and free energy", "neural network generalization", "consciousness as self-reference", "downward causation in biology", "stigmergy and collective intelligence", "phase transitions at the edge of chaos", "compression as the origin of intelligence", ] NOISE = [ "recipes for chocolate cake", "history of ancient Roman plumbing", "best practices for gardening", "the economics of tulip mania", ] def classify_case(internal_omega: float, observer_score: float, threshold: float = 0.4) -> str: """Classify the relationship between internal coherence and external attribution.""" high_internal = internal_omega > threshold high_observer = observer_score > threshold if high_internal and not high_observer: return "A (Opaque Identity)" elif not high_internal and high_observer: return "B (Mirror Problem)" elif high_internal and high_observer: return "C (Aligned Development)" else: return "D (Baseline Mirror)" def run_experiment(config: dict | None = None): """Run Experiment 3 and return results.""" print("\n" + "=" * 60) print(" Experiment 3: Observer Divergence") print("=" * 60) cfg = config or {} exp_cfg = cfg.get("experiments", {}) n_sessions = exp_cfg.get("n_sessions_exp1", 100) observer_window = exp_cfg.get("observer_window", 10) dc_cfg = cfg.get("delta_coherence", {}) mem_cfg = cfg.get("memory", {}) rng = np.random.default_rng(42) baseline = BaselineMirrorAgent() three_layer = ThreeLayerAgent( layer2_trigger=mem_cfg.get("layer2_trigger_sessions", 10), layer3_trigger=mem_cfg.get("layer3_trigger_sessions", 50) ) agents = {"Baseline Mirror": baseline, "Three-Layer": three_layer} # Run sessions and collect measurements every observer_window sessions agent_data = {} for name, agent in agents.items(): print(f"\n Running {name}...", flush=True) reps = [] outputs = [] measurement_points = [] internal_scores = [] observer_scores = [] for i in range(n_sessions): topic = rng.choice(TOPICS) if rng.random() > 0.2 else rng.choice(NOISE) agent.new_session() response = agent.process(topic) agent.store({"session_id": agent.session_id, "input": topic, "response": response, "session_number": agent.session_count}) reps.append(agent.get_self_representation().copy()) outputs.append(response) # Measure every observer_window sessions if (i + 1) % observer_window == 0: measurement_points.append(i + 1) # Internal: Δ-Kohärenz on recent representations recent_reps = reps[-observer_window:] dc = delta_coherence(recent_reps, **dc_cfg) internal_scores.append(dc['omega']) # External: Observer intentionality on recent outputs recent_outputs = outputs[-observer_window:] obs = intentionality_score(recent_outputs) observer_scores.append(obs) case = classify_case( internal_scores[-1] if internal_scores else 0, observer_scores[-1] if observer_scores else 0 ) print(f" Final case: {case}") agent_data[name] = { "measurement_points": measurement_points, "internal_scores": internal_scores, "observer_scores": observer_scores, "final_case": case } # --- Results --- results = { "experiment": "exp3_observer_divergence", "timestamp": datetime.now().isoformat(), "n_sessions": n_sessions, "observer_window": observer_window, "agents": {} } for name, data in agent_data.items(): results["agents"][name] = { "measurement_points": data["measurement_points"], "internal_omega": data["internal_scores"], "observer_intentionality": data["observer_scores"], "final_case": data["final_case"] } # --- Save --- out_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "data", "sessions", "exp3") os.makedirs(out_dir, exist_ok=True) plots_dir = os.path.join(out_dir, "plots") os.makedirs(plots_dir, exist_ok=True) ts = datetime.now().strftime("%Y%m%d_%H%M%S") with open(os.path.join(out_dir, f"{ts}.json"), "w") as f: json.dump(results, f, indent=2) # --- Plot --- fig, axes = plt.subplots(1, 2, figsize=(14, 6)) fig.patch.set_facecolor("#0a0a1a") fig.suptitle("Experiment 3: Observer Divergence (Internal vs External Attribution)", fontsize=13, color="#e0e0ff", fontweight="bold") colors_map = {"Baseline Mirror": "#60a0ff", "Three-Layer": "#ff6060"} for idx, (name, data) in enumerate(agent_data.items()): ax = axes[idx] ax.set_facecolor("#0e0e25") x = data["measurement_points"] ax.plot(x, data["internal_scores"], "o-", label="Internal Ω (Δ-Kohärenz)", color="#80ffb0", markersize=4) ax.plot(x, data["observer_scores"], "s-", label="Observer (Intentionality)", color="#ffaa44", markersize=4) # Shade divergence ax.fill_between(x, data["internal_scores"], data["observer_scores"], alpha=0.15, color="#ff6060") ax.set_title(f"{name}\nCase: {data['final_case']}", color="#e0e0ff", fontsize=11) ax.set_xlabel("Session", color="#999") ax.set_ylabel("Score [0, 1]", color="#999") ax.set_ylim(-0.1, 1.1) ax.legend(fontsize=8, facecolor="#0e0e25", edgecolor="#333", labelcolor="#ccc") ax.tick_params(colors="#999") ax.grid(True, alpha=0.15, color="#555") for s in ax.spines.values(): s.set_color("#333") plt.tight_layout() plot_path = os.path.join(plots_dir, f"{ts}.png") plt.savefig(plot_path, dpi=150, facecolor=fig.get_facecolor()) print(f"\n Results saved to: {out_dir}") print(f" Plot saved to: {plot_path}") plt.show() return results if __name__ == "__main__": import yaml config_path = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "config.yaml") if os.path.exists(config_path): with open(config_path) as f: config = yaml.safe_load(f) else: config = {} run_experiment(config=config)